Rubin NVL72 platform
Search documents
英伟达发布下一代 Rubin 平台-AI industry update - Nvidia announces next-generation Rubin platform
2026-01-13 11:56
Summary of Key Points from the Conference Call Industry and Company Overview - The conference call focuses on the **AI industry**, specifically highlighting **Nvidia** and its next-generation **Rubin platform** announced at **CES 2026** [1][2]. Core Insights and Arguments - **Architecture Changes**: Nvidia's Rubin platform introduces a **Context Memory Storage Platform** that utilizes **Ethernet switches** for external storage, indicating a shift from reliance on high-bandwidth memory (HBM) to dedicated storage racks [1][2]. - **Power Consumption**: The Rubin NVL72 system reportedly consumes **240kW**, which is **twice** the power of the previous Blackwell generation. This increase in power consumption is attributed to the use of **400G SerDes technology** and may lead to higher demand for capacitors in power supplies [3][5]. - **Cooling Efficiency**: The Rubin NVL72 employs a **fully water-cooled system** without fan motors, which is a departure from the Blackwell system. The cooling liquid operates at **45°C**, which is crucial for maintaining efficiency [4]. - **Data Center Bottlenecks**: There are concerns that power supply will become a bottleneck for data centers in the US, especially with the anticipated increase in power consumption from future platforms. Solutions such as **natural gas-fired generators** and **storage batteries** are deemed essential [5]. Additional Important Points - **Memory Pooling**: The architecture's connection to external storage via Ethernet switches resembles memory pooling techniques used in the **CXL standard**, indicating a broader trend in hardware design [2]. - **Material Innovations**: New materials such as **fluoroplastic substrates** and **high-grade copper foil** may be used in the PCB designs to mitigate signal degradation, which is critical for high-speed signaling [3]. - **Market Implications**: The changes in Nvidia's architecture and power consumption patterns could have significant implications for the **electronic parts industry**, particularly in the demand for capacitors and cooling solutions [3][5]. This summary encapsulates the key points discussed in the conference call, providing insights into Nvidia's strategic direction and its potential impact on the AI and electronic parts industries.